libcitadel/lib/mime_parser.c

   1 /*
   2  * This is the MIME parser for Citadel.
   3  *
   4  * Copyright (c) 1998-2010 by the citadel.org development team.
   5  * This code is distributed under the GNU General Public License v3.
   6  *
   7  */
   8
   9 #include <stdlib.h>
  10 #include <unistd.h>
  11 #include <stdio.h>
  12 #include <signal.h>
  13 #include <sys/types.h>
  14 #include <ctype.h>
  15 #include <string.h>
  16 #include <sys/stat.h>
  17 #include <sys/types.h>
  18 #include <dirent.h>
  19 #include <errno.h>
  20
  21 #include "xdgmime/xdgmime.h"
  22 #include "libcitadel.h"
  23 #include "libcitadellocal.h"
  24
  25 const unsigned char FromHexTable [256] = {
  26         0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, //  0
  27         0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, // 10
  28         0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, // 20
  29         0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, // 30
  30         0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0x00, 0x01, // 40
  31         0x02, 0x03, 0x04, 0x05, 0x06, 0x07, 0x08, 0x09, 0xFF, 0xFF, // 50
  32         0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0x0A, 0x0B, 0x0C, 0x0D, 0x0E, // 60
  33         0x0F, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, // 70
  34         0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, // 80
  35         0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0x0A, 0x0B, 0x0C, // 90
  36         0x0D, 0x0E, 0x0F, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, //100
  37         0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, //110
  38         0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, //120
  39         0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, //130
  40         0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, //140
  41         0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, //150
  42         0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, //160
  43         0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, //170
  44         0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, //180
  45         0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, //190
  46         0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, //200
  47         0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, //210
  48         0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, //220
  49         0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, //230
  50         0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, //240
  51         0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF                          //250
  52 };
  53
  54
  55 long extract_key(char *target, char *source, long sourcelen, char *key, long keylen, char KeyEnd)
  56 {
  57         char *sptr, *ptr = NULL;
  58         int double_quotes = 0;
  59         long RealKeyLen = keylen;
  60
  61         sptr = source;
  62
  63         while (sptr != NULL)
  64         {
  65                 ptr = bmstrcasestr_len(sptr, sourcelen - (sptr - source),
  66                                        key, keylen);
  67                 if(ptr != NULL)
  68                 {
  69                         while (isspace(*(ptr + RealKeyLen)))
  70                                 RealKeyLen ++;
  71                         if (*(ptr + RealKeyLen) == KeyEnd)
  72                         {
  73                                 sptr = NULL;
  74                                 RealKeyLen ++;
  75                         }
  76                         else
  77                         {
  78                                 sptr = ptr + RealKeyLen + 1;
  79                         }
  80                 }
  81                 else
  82                         sptr = ptr;
  83         }
  84         if (ptr == NULL) {
  85                 *target = '\0';
  86                 return 0;
  87         }
  88         strcpy(target, (ptr + RealKeyLen));
  89
  90         for (ptr=target; (*ptr != 0); ptr++) {
  91
  92                 /* A semicolon means we've hit the end of the key, unless we're inside double quotes */
  93                 if ( (double_quotes != 1) && (*ptr == ';')) {
  94                         *ptr = 0;
  95                 }
  96
  97                 /* if we find double quotes, we've got a great set of string boundaries */
  98                 if (*ptr == '\"') {
  99                         ++double_quotes;
 100                         if (double_quotes == 1) {
 101                                 strcpy(ptr, ptr+1);
 102                         }
 103                         else {
 104                                 *ptr = 0;
 105                         }
 106                 }
 107         }
 108         *ptr = '\0';
 109         return ptr - target;
 110 }
 111
 112
 113 /*
 114  * For non-multipart messages, we need to generate a quickie partnum of "1"
 115  * to return to callback functions.  Some callbacks demand it.
 116  */
 117 char *fixed_partnum(char *supplied_partnum) {
 118         if (supplied_partnum == NULL) return "1";
 119         if (strlen(supplied_partnum)==0) return "1";
 120         return supplied_partnum;
 121 }
 122
 123
 124 static inline unsigned int _decode_hex(const char *Source)
 125 {
 126         int ret = '?';
 127         unsigned char LO_NIBBLE;
 128         unsigned char HI_NIBBLE;
 129
 130         HI_NIBBLE = FromHexTable[(unsigned char) *Source];
 131         LO_NIBBLE = FromHexTable[(unsigned char) *(Source+1)];
 132
 133         if ((LO_NIBBLE == 0xFF) || (LO_NIBBLE == 0xFF))
 134                 return ret;
 135         ret = HI_NIBBLE;
 136         ret = ret << 4;
 137         ret = ret | LO_NIBBLE;
 138         return ret;
 139 }
 140
 141 unsigned int decode_hex(char *Source) {return _decode_hex(Source);}
 142
 143 /*
 144  * Convert "quoted-printable" to binary.  Returns number of bytes decoded.
 145  * according to RFC2045 section 6.7
 146  */
 147 int CtdlDecodeQuotedPrintable(char *decoded, char *encoded, int sourcelen) {
 148         unsigned int ch;
 149         int decoded_length = 0;
 150         int pos = 0;
 151
 152         while (pos < sourcelen)
 153         {
 154                 if (*(encoded + pos) == '=')
 155                 {
 156                         pos ++;
 157                         if (*(encoded + pos) == '\n')
 158                         {
 159                                 pos ++;
 160                         }
 161                         else if (*(encoded + pos) == '\r')
 162                         {
 163                                 pos ++;
 164                                 if (*(encoded + pos) == '\n')
 165                                         pos++;
 166                         }
 167                         else
 168                         {
 169                                 ch = 0;
 170                                 ch = _decode_hex(&encoded[pos]);
 171                                 pos += 2;
 172                                 decoded[decoded_length++] = ch;
 173                         }
 174                 }
 175                 else
 176                 {
 177                         decoded[decoded_length++] = encoded[pos];
 178                         pos += 1;
 179                 }
 180         }
 181         decoded[decoded_length] = 0;
 182         return(decoded_length);
 183 }
 184
 185
 186 /*
 187  * Given a message or message-part body and a length, handle any necessary
 188  * decoding and pass the request up the stack.
 189  */
 190 void mime_decode(char *partnum,
 191                  char *part_start, size_t length,
 192                  char *content_type, char *charset, char *encoding,
 193                  char *disposition,
 194                  char *id,
 195                  char *name, char *filename,
 196                  MimeParserCallBackType CallBack,
 197                  MimeParserCallBackType PreMultiPartCallBack,
 198                  MimeParserCallBackType PostMultiPartCallBack,
 199                  void *userdata,
 200                  int dont_decode)
 201 {
 202
 203         char *decoded;
 204         size_t bytes_decoded = 0;
 205
 206         /* Some encodings aren't really encodings */
 207         if (!strcasecmp(encoding, "7bit"))
 208                 strcpy(encoding, "");
 209         if (!strcasecmp(encoding, "8bit"))
 210                 strcpy(encoding, "");
 211         if (!strcasecmp(encoding, "binary"))
 212                 strcpy(encoding, "");
 213
 214         /* If this part is not encoded, send as-is */
 215         if ( (strlen(encoding) == 0) || (dont_decode)) {
 216                 if (CallBack != NULL) {
 217                         CallBack(name,
 218                                  filename,
 219                                  fixed_partnum(partnum),
 220                                  disposition,
 221                                  part_start,
 222                                  content_type,
 223                                  charset,
 224                                  length,
 225                                  encoding,
 226                                  id,
 227                                  userdata);
 228                         }
 229                 return;
 230         }
 231
 232         /* Fail silently if we hit an unknown encoding. */
 233         if ((strcasecmp(encoding, "base64"))
 234             && (strcasecmp(encoding, "quoted-printable"))) {
 235                 return;
 236         }
 237
 238         /*
 239          * Allocate a buffer for the decoded data.  The output buffer is slightly
 240          * larger than the input buffer; this assumes that the decoded data
 241          * will never be significantly larger than the encoded data.  This is a
 242          * safe assumption with base64, uuencode, and quoted-printable.
 243          */
 244         decoded = malloc(length + 32768);
 245         if (decoded == NULL) {
 246                 return;
 247         }
 248
 249         if (!strcasecmp(encoding, "base64")) {
 250                 bytes_decoded = CtdlDecodeBase64(decoded, part_start, length);
 251         }
 252         else if (!strcasecmp(encoding, "quoted-printable")) {
 253                 bytes_decoded = CtdlDecodeQuotedPrintable(decoded, part_start, length);
 254         }
 255
 256         if (bytes_decoded > 0) if (CallBack != NULL) {
 257                         char encoding_buf[SIZ];
 258
 259                         strcpy(encoding_buf, "binary");
 260                         CallBack(name,
 261                                  filename,
 262                                  fixed_partnum(partnum),
 263                                  disposition,
 264                                  decoded,
 265                                  content_type,
 266                                  charset,
 267                                  bytes_decoded,
 268                                  encoding_buf,
 269                                  id,
 270                                  userdata);
 271         }
 272
 273         free(decoded);
 274 }
 275
 276 /*
 277  * this is the extract of mime_decode which can be called if 'dont_decode' was set;
 278  * to save the cpu intense process of decoding to the time when it realy wants the content.
 279  * returns:
 280  *   - > 0 we decoded something, its on *decoded, you need to free it.
 281  *   - = 0 no need to decode stuff. *decoded will be NULL.
 282  *   - < 0 an error occured, either an unknown encoding, or alloc failed. no need to free.
 283  */
 284 int mime_decode_now (char *part_start,
 285                      size_t length,
 286                      char *encoding,
 287                      char **decoded,
 288                      size_t *bytes_decoded)
 289 {
 290         *bytes_decoded = 0;
 291         *decoded = NULL;
 292         /* Some encodings aren't really encodings */
 293         if (!strcasecmp(encoding, "7bit"))
 294                 strcpy(encoding, "");
 295         if (!strcasecmp(encoding, "8bit"))
 296                 strcpy(encoding, "");
 297         if (!strcasecmp(encoding, "binary"))
 298                 strcpy(encoding, "");
 299
 300         /* If this part is not encoded, send as-is */
 301         if (strlen(encoding) == 0) {
 302                 return 0;
 303         }
 304
 305
 306         /* Fail if we hit an unknown encoding. */
 307         if ((strcasecmp(encoding, "base64"))
 308             && (strcasecmp(encoding, "quoted-printable"))) {
 309                 return -1;
 310         }
 311
 312         /*
 313          * Allocate a buffer for the decoded data.  The output buffer is slightly
 314          * larger than the input buffer; this assumes that the decoded data
 315          * will never be significantly larger than the encoded data.  This is a
 316          * safe assumption with base64, uuencode, and quoted-printable.
 317          */
 318         *decoded = malloc(length + 32768);
 319         if (decoded == NULL) {
 320                 return -1;
 321         }
 322
 323         if (!strcasecmp(encoding, "base64")) {
 324                 *bytes_decoded = CtdlDecodeBase64(*decoded, part_start, length);
 325                 return 1;
 326         }
 327         else if (!strcasecmp(encoding, "quoted-printable")) {
 328                 *bytes_decoded = CtdlDecodeQuotedPrintable(*decoded, part_start, length);
 329                 return 1;
 330         }
 331         return -1;
 332 }
 333
 334 typedef enum _eIntMimeHdrs {
 335         boundary,
 336         startary,
 337         endary,
 338         content_type,
 339         charset,
 340         encoding,
 341         content_type_name,
 342         content_disposition_name,
 343         filename,
 344         disposition,
 345         id,
 346         eMax /* don't move ! */
 347 } eIntMimeHdrs;
 348
 349 typedef struct _CBufStr {
 350         char Key[SIZ];
 351         long len;
 352 }CBufStr;
 353
 354 typedef struct _interesting_mime_headers {
 355         CBufStr b[eMax];
 356         long content_length;
 357         long is_multipart;
 358 } interesting_mime_headers;
 359
 360
 361 static void FlushInterestingMimes(interesting_mime_headers *m)
 362 {
 363         int i;
 364
 365         for (i = 0; i < eMax; i++) {
 366              m->b[i].Key[0] = '\0';
 367              m->b[i].len = 0;
 368         }
 369         m->content_length = -1;
 370 }
 371 static interesting_mime_headers *InitInterestingMimes(void)
 372 {
 373         interesting_mime_headers *m;
 374         m = (interesting_mime_headers*) malloc( sizeof(interesting_mime_headers));
 375
 376         FlushInterestingMimes(m);
 377
 378         return m;
 379 }
 380
 381
 382 static long parse_MimeHeaders(interesting_mime_headers *m,
 383                               char** pcontent_start,
 384                               char *content_end)
 385 {
 386         char buf[SIZ];
 387         char header[SIZ];
 388         long headerlen;
 389         char *ptr, *pch;
 390         int buflen = 0;
 391         int i;
 392
 393         /* Learn interesting things from the headers */
 394         ptr = *pcontent_start;
 395         *header = '\0';
 396         headerlen = 0;
 397         do {
 398                 ptr = memreadlinelen(ptr, buf, SIZ, &buflen);
 399
 400                 for (i = 0; i < buflen; ++i) {
 401                         if (isspace(buf[i])) {
 402                                 buf[i] = ' ';
 403                         }
 404                 }
 405
 406                 if (!isspace(buf[0]) && (headerlen > 0)) {
 407                         if (!strncasecmp(header, "Content-type:", 13)) {
 408                                 memcpy (m->b[content_type].Key, &header[13], headerlen - 12);
 409                                 m->b[content_type].len = striplt (m->b[content_type].Key);
 410
 411                                 m->b[content_type_name].len = extract_key(m->b[content_type_name].Key, CKEY(m->b[content_type]), HKEY("name"), '=');
 412                                 m->b[charset].len           = extract_key(m->b[charset].Key,           CKEY(m->b[content_type]), HKEY("charset"), '=');
 413                                 m->b[boundary].len          = extract_key(m->b[boundary].Key,          header,       headerlen,  HKEY("boundary"), '=');
 414
 415                                 /* Deal with weird headers */
 416                                 pch = strchr(m->b[content_type].Key, ' ');
 417                                 if (pch != NULL) {
 418                                         *pch = '\0';
 419                                         m->b[content_type].len = m->b[content_type].Key - pch;
 420                                 }
 421                                 pch = strchr(m->b[content_type].Key, ';');
 422                                 if (pch != NULL) {
 423                                         *pch = '\0';
 424                                         m->b[content_type].len = m->b[content_type].Key - pch;
 425                                 }
 426                         }
 427                         else if (!strncasecmp(header, "Content-Disposition:", 20)) {
 428                                 memcpy (m->b[disposition].Key, &header[20], headerlen - 19);
 429                                 m->b[disposition].len = striplt(m->b[disposition].Key);
 430
 431                                 m->b[content_disposition_name].len = extract_key(m->b[content_disposition_name].Key, CKEY(m->b[disposition]), HKEY("name"), '=');
 432                                 m->b[filename].len                 = extract_key(m->b[filename].Key,                 CKEY(m->b[disposition]), HKEY("filename"), '=');
 433                                 pch = strchr(m->b[disposition].Key, ';');
 434                                 if (pch != NULL) *pch = '\0';
 435                                 m->b[disposition].len = striplt(m->b[disposition].Key);
 436                         }
 437                         else if (!strncasecmp(header, "Content-ID:", 11)) {
 438                                 memcpy(m->b[id].Key, &header[11], headerlen);
 439                                 striplt(m->b[id].Key);
 440                                 m->b[id].len = stripallbut(m->b[id].Key, '<', '>');
 441                         }
 442                         else if (!strncasecmp(header, "Content-length: ", 15)) {
 443                                 char *clbuf;
 444                                 clbuf = &header[15];
 445                                 while (isspace(*clbuf))
 446                                         clbuf ++;
 447                                 m->content_length = (size_t) atol(clbuf);
 448                         }
 449                         else if (!strncasecmp(header, "Content-transfer-encoding: ", 26)) {
 450                                 memcpy(m->b[encoding].Key, &header[26], headerlen - 26);
 451                                 m->b[encoding].len = striplt(m->b[encoding].Key);
 452                         }
 453                         *header = '\0';
 454                         headerlen = 0;
 455                 }
 456                 if ((headerlen + buflen + 2) < SIZ) {
 457                         memcpy(&header[headerlen], buf, buflen);
 458                         headerlen += buflen;
 459                         header[headerlen] = '\0';
 460                 }
 461                 if (ptr >= content_end) {
 462                         return -1;
 463                 }
 464         } while ((!IsEmptyStr(buf)) && (*ptr != 0));
 465
 466         m->is_multipart = m->b[boundary].len != 0;
 467         *pcontent_start = ptr;
 468
 469         return 0;
 470 }
 471
 472
 473 static int IsAsciiEncoding(interesting_mime_headers *m)
 474 {
 475
 476         if ((m->b[encoding].len != 0) &&
 477             (strcasecmp(m->b[encoding].Key, "base64") == 0))
 478                 return 1;
 479         if ((m->b[encoding].len != 0) &&
 480             (strcmp(m->b[encoding].Key, "quoted-printable") == 0))
 481                 return 1;
 482
 483         return 0;
 484 }
 485
 486 static char *FindNextContent(char *ptr,
 487                              char *content_end,
 488                              interesting_mime_headers *SubMimeHeaders,
 489                              interesting_mime_headers *m)
 490 {
 491         char *next_boundary;
 492         char  tmp;
 493
 494         if (IsAsciiEncoding(SubMimeHeaders)) {
 495                 tmp = *content_end;
 496                 *content_end = '\0';
 497
 498                 /**
 499                  * ok, if we have a content length of the mime part,
 500                  * try skipping the content on the search for the next
 501                  * boundary. since we don't trust the content_length
 502                  * to be all accurate, and suspect it to lose one digit
 503                  * per line with a line length of 80 chars, we need
 504                  * to start searching a little before..
 505                  */
 506
 507                 if ((SubMimeHeaders->content_length != -1) &&
 508                     (SubMimeHeaders->content_length > 10))
 509                 {
 510                         char *pptr;
 511                         long lines;
 512
 513                         lines = SubMimeHeaders->content_length / 80;
 514                         pptr = ptr + SubMimeHeaders->content_length - lines - 10;
 515                         if (pptr < content_end)
 516                                 ptr = pptr;
 517                 }
 518
 519                 next_boundary = strstr(ptr, m->b[startary].Key);
 520                 *content_end = tmp;
 521         }
 522         else {
 523                 char *srch;
 524                 /**
 525                  * ok, if we have a content length of the mime part,
 526                  * try skipping the content on the search for the next
 527                  * boundary. since we don't trust the content_length
 528                  * to be all accurate, start searching a little before..
 529                  */
 530
 531                 if ((SubMimeHeaders->content_length != -1) &&
 532                     (SubMimeHeaders->content_length > 10))
 533                 {
 534                         char *pptr;
 535                         pptr = ptr + SubMimeHeaders->content_length - 10;
 536                         if (pptr < content_end)
 537                                 ptr = pptr;
 538                 }
 539                 next_boundary = NULL;
 540                 for (srch=ptr; srch<content_end; ++srch) {
 541                         if (!memcmp(srch,
 542                                     m->b[startary].Key,
 543                                     m->b[startary].len))
 544                         {
 545                                 next_boundary = srch;
 546                                 srch = content_end;
 547                         }
 548                 }
 549
 550         }
 551         return next_boundary;
 552 }
 553
 554 /*
 555  * Break out the components of a multipart message
 556  * (This function expects to be fed HEADERS + CONTENT)
 557  * Note: NULL can be supplied as content_end; in this case, the message is
 558  * considered to have ended when the parser encounters a 0x00 byte.
 559  */
 560 static void recurseable_mime_parser(char *partnum,
 561                                     char *content_start, char *content_end,
 562                                     MimeParserCallBackType CallBack,
 563                                     MimeParserCallBackType PreMultiPartCallBack,
 564                                     MimeParserCallBackType PostMultiPartCallBack,
 565                                     void *userdata,
 566                                     int dont_decode,
 567                                     interesting_mime_headers *m)
 568 {
 569         interesting_mime_headers *SubMimeHeaders;
 570         char     *ptr;
 571         char     *part_start;
 572         char     *part_end = NULL;
 573         char     *evaluate_crlf_ptr = NULL;
 574         char     *next_boundary;
 575         char      nested_partnum[256];
 576         int       crlf_in_use = 0;
 577         int       part_seq = 0;
 578         CBufStr  *chosen_name;
 579
 580
 581         /* If this is a multipart message, then recursively process it */
 582         ptr = content_start;
 583         part_start = NULL;
 584         if (m->is_multipart) {
 585
 586                 /* Tell the client about this message's multipartedness */
 587                 if (PreMultiPartCallBack != NULL) {
 588                         PreMultiPartCallBack("",
 589                                              "",
 590                                              partnum,
 591                                              "",
 592                                              NULL,
 593                                              m->b[content_type].Key,
 594                                              m->b[charset].Key,
 595                                              0,
 596                                              m->b[encoding].Key,
 597                                              m->b[id].Key,
 598                                              userdata);
 599                 }
 600
 601                 /* Figure out where the boundaries are */
 602                 m->b[startary].len = snprintf(m->b[startary].Key, SIZ, "--%s", m->b[boundary].Key);
 603                 SubMimeHeaders = InitInterestingMimes ();
 604                 if (*ptr == '\r')
 605                         ptr ++;
 606                 if (*ptr == '\n')
 607                         ptr ++;
 608                 if (strncmp(ptr, m->b[startary].Key, m->b[startary].len) == 0)
 609                         ptr += m->b[startary].len;
 610                 if (*ptr == '\r')
 611                         ptr ++;
 612                 if (*ptr == '\n')
 613                         ptr ++;
 614                 part_start = NULL;
 615                 do {
 616
 617                         if (parse_MimeHeaders(SubMimeHeaders, &ptr, content_end) != 0)
 618                                 break;
 619                         part_start = ptr;
 620
 621                         next_boundary = FindNextContent(ptr,
 622                                                         content_end,
 623                                                         SubMimeHeaders,
 624                                                         m);
 625                         if ((next_boundary != NULL) &&
 626                             (next_boundary - part_start < 3))
 627                                 continue;
 628
 629                         if ( (part_start != NULL) && (next_boundary != NULL) ) {
 630                                 part_end = next_boundary;
 631                                 --part_end;             /* omit the trailing LF */
 632                                 if (crlf_in_use) {
 633                                         --part_end;     /* omit the trailing CR */
 634                                 }
 635
 636                                 if (!IsEmptyStr(partnum)) {
 637                                         snprintf(nested_partnum,
 638                                                  sizeof nested_partnum,
 639                                                  "%s.%d", partnum,
 640                                                  ++part_seq);
 641                                 }
 642                                 else {
 643                                         snprintf(nested_partnum,
 644                                                  sizeof nested_partnum,
 645                                                  "%d", ++part_seq);
 646                                 }
 647                                 recurseable_mime_parser(nested_partnum,
 648                                                         part_start,
 649                                                         part_end,
 650                                                         CallBack,
 651                                                         PreMultiPartCallBack,
 652                                                         PostMultiPartCallBack,
 653                                                         userdata,
 654                                                         dont_decode,
 655                                                         SubMimeHeaders);
 656                         }
 657
 658                         if (next_boundary != NULL) {
 659                                 /* If we pass out of scope, don't attempt to
 660                                  * read past the end boundary. */
 661                                 if ((*(next_boundary + m->b[startary].len + 1) == '-') &&
 662                                     (*(next_boundary + m->b[startary].len + 2) == '-') ){
 663                                         ptr = content_end;
 664                                 }
 665                                 else {
 666                                         /* Set up for the next part. */
 667                                         part_start = strstr(next_boundary, "\n");
 668
 669                                         /* Determine whether newlines are LF or CRLF */
 670                                         evaluate_crlf_ptr = part_start;
 671                                         --evaluate_crlf_ptr;
 672                                         if ((*evaluate_crlf_ptr == '\r') &&
 673                                             (*(evaluate_crlf_ptr + 1) == '\n'))
 674                                         {
 675                                                 crlf_in_use = 1;
 676                                         }
 677                                         else {
 678                                                 crlf_in_use = 0;
 679                                         }
 680
 681                                         /* Advance past the LF ... now we're in the next part */
 682                                         ++part_start;
 683                                         ptr = part_start;
 684                                 }
 685                         }
 686                         else {
 687                                 /* Invalid end of multipart.  Bail out! */
 688                                 ptr = content_end;
 689                         }
 690                         FlushInterestingMimes(SubMimeHeaders);
 691                 } while ( (ptr < content_end) && (next_boundary != NULL) );
 692
 693                 free(SubMimeHeaders);
 694
 695                 if (PostMultiPartCallBack != NULL) {
 696                         PostMultiPartCallBack("",
 697                                               "",
 698                                               partnum,
 699                                               "",
 700                                               NULL,
 701                                               m->b[content_type].Key,
 702                                               m->b[charset].Key,
 703                                               0,
 704                                               m->b[encoding].Key,
 705                                               m->b[id].Key,
 706                                               userdata);
 707                 }
 708         } /* If it's not a multipart message, then do something with it */
 709         else {
 710                 size_t length;
 711                 part_start = ptr;
 712                 length = content_end - part_start;
 713                 ptr = part_end = content_end;
 714
 715
 716                 /* The following code will truncate the MIME part to the size
 717                  * specified by the Content-length: header.   We have commented it
 718                  * out because these headers have a tendency to be wrong.
 719                  *
 720                  *      if ( (content_length > 0) && (length > content_length) ) {
 721                  *              length = content_length;
 722                  *      }
 723                  */
 724
 725                 /* Sometimes the "name" field is tacked on to Content-type,
 726                  * and sometimes it's tacked on to Content-disposition.  Use
 727                  * whichever one we have.
 728                  */
 729                 if (m->b[content_disposition_name].len > m->b[content_type_name].len) {
 730                         chosen_name = &m->b[content_disposition_name];
 731                 }
 732                 else {
 733                         chosen_name = &m->b[content_type_name];
 734                 }
 735
 736                 /* Ok, we've got a non-multipart part here, so do something with it.
 737                  */
 738                 mime_decode(partnum,
 739                             part_start,
 740                             length,
 741                             m->b[content_type].Key,
 742                             m->b[charset].Key,
 743                             m->b[encoding].Key,
 744                             m->b[disposition].Key,
 745                             m->b[id].Key,
 746                             chosen_name->Key,
 747                             m->b[filename].Key,
 748                             CallBack,
 749                             NULL, NULL,
 750                             userdata,
 751                             dont_decode
 752                         );
 753
 754                 /*
 755                  * Now if it's an encapsulated message/rfc822 then we have to recurse into it
 756                  */
 757                 if (!strcasecmp(&m->b[content_type].Key[0], "message/rfc822")) {
 758
 759                         if (PreMultiPartCallBack != NULL) {
 760                                 PreMultiPartCallBack("",
 761                                                      "",
 762                                                      partnum,
 763                                                      "",
 764                                                      NULL,
 765                                                      m->b[content_type].Key,
 766                                                      m->b[charset].Key,
 767                                                      0,
 768                                                      m->b[encoding].Key,
 769                                                      m->b[id].Key,
 770                                                      userdata);
 771                         }
 772                         if (CallBack != NULL) {
 773                                 if (strlen(partnum) > 0) {
 774                                         snprintf(nested_partnum,
 775                                                  sizeof nested_partnum,
 776                                                  "%s.%d", partnum,
 777                                                  ++part_seq);
 778                                 }
 779                                 else {
 780                                         snprintf(nested_partnum,
 781                                                  sizeof nested_partnum,
 782                                                  "%d", ++part_seq);
 783                                 }
 784                                 the_mime_parser(nested_partnum,
 785                                                 part_start,
 786                                                 part_end,
 787                                                 CallBack,
 788                                                 PreMultiPartCallBack,
 789                                                 PostMultiPartCallBack,
 790                                                 userdata,
 791                                                 dont_decode
 792                                         );
 793                         }
 794                         if (PostMultiPartCallBack != NULL) {
 795                                 PostMultiPartCallBack("",
 796                                                       "",
 797                                                       partnum,
 798                                                       "",
 799                                                       NULL,
 800                                                       m->b[content_type].Key,
 801                                                       m->b[charset].Key,
 802                                                       0,
 803                                                       m->b[encoding].Key,
 804                                                       m->b[id].Key,
 805                                                       userdata);
 806                         }
 807
 808
 809                 }
 810
 811         }
 812
 813 }
 814
 815 /*
 816  * Break out the components of a multipart message
 817  * (This function expects to be fed HEADERS + CONTENT)
 818  * Note: NULL can be supplied as content_end; in this case, the message is
 819  * considered to have ended when the parser encounters a 0x00 byte.
 820  */
 821 void the_mime_parser(char *partnum,
 822                      char *content_start, char *content_end,
 823                      MimeParserCallBackType CallBack,
 824                      MimeParserCallBackType PreMultiPartCallBack,
 825                      MimeParserCallBackType PostMultiPartCallBack,
 826                      void *userdata,
 827                      int dont_decode)
 828 {
 829         interesting_mime_headers *m;
 830
 831         /* If the caller didn't supply an endpointer, generate one by measure */
 832         if (content_end == NULL) {
 833                 content_end = &content_start[strlen(content_start)];
 834         }
 835
 836         m = InitInterestingMimes();
 837
 838         if (!parse_MimeHeaders(m, &content_start, content_end))
 839         {
 840
 841                 recurseable_mime_parser(partnum,
 842                                         content_start, content_end,
 843                                         CallBack,
 844                                         PreMultiPartCallBack,
 845                                         PostMultiPartCallBack,
 846                                         userdata,
 847                                         dont_decode,
 848                                         m);
 849         }
 850         free(m);
 851 }
 852
 853 /*
 854  * Entry point for the MIME parser.
 855  * (This function expects to be fed HEADERS + CONTENT)
 856  * Note: NULL can be supplied as content_end; in this case, the message is
 857  * considered to have ended when the parser encounters a 0x00 byte.
 858  */
 859 void mime_parser(char *content_start,
 860                  char *content_end,
 861                  MimeParserCallBackType CallBack,
 862                  MimeParserCallBackType PreMultiPartCallBack,
 863                  MimeParserCallBackType PostMultiPartCallBack,
 864                  void *userdata,
 865                  int dont_decode)
 866 {
 867
 868         the_mime_parser("", content_start, content_end,
 869                         CallBack,
 870                         PreMultiPartCallBack,
 871                         PostMultiPartCallBack,
 872                         userdata, dont_decode);
 873 }
 874
 875
 876
 877
 878
 879
 880 typedef struct _MimeGuess {
 881         const char *Pattern;
 882         size_t PatternLen;
 883         long PatternOffset;
 884         const char *MimeString;
 885 } MimeGuess;
 886
 887 MimeGuess MyMimes [] = {
 888         {
 889                 "GIF",
 890                 3,
 891                 0,
 892                 "image/gif"
 893         },
 894         {
 895                 "\xff\xd8",
 896                 2,
 897                 0,
 898                 "image/jpeg"
 899         },
 900         {
 901                 "\x89PNG",
 902                 4,
 903                 0,
 904                 "image/png"
 905         },
 906         { // last...
 907                 "",
 908                 0,
 909                 0,
 910                 ""
 911         }
 912 };
 913
 914
 915 const char *GuessMimeType(const char *data, size_t dlen)
 916 {
 917         int MimeIndex = 0;
 918
 919         while (MyMimes[MimeIndex].PatternLen != 0)
 920         {
 921                 if ((MyMimes[MimeIndex].PatternLen +
 922                      MyMimes[MimeIndex].PatternOffset < dlen) &&
 923                     strncmp(MyMimes[MimeIndex].Pattern,
 924                             &data[MyMimes[MimeIndex].PatternOffset],
 925                             MyMimes[MimeIndex].PatternLen) == 0)
 926                 {
 927                         return MyMimes[MimeIndex].MimeString;
 928                 }
 929                 MimeIndex ++;
 930         }
 931         /*
 932          * ok, our simple minded algorythm didn't find anything,
 933          * let the big chegger try it, he wil default to application/octet-stream
 934          */
 935         return (xdg_mime_get_mime_type_for_data(data, dlen));
 936 }
 937
 938
 939 const char* GuessMimeByFilename(const char *what, size_t len)
 940 {
 941         /* we know some hardcoded on our own, try them... */
 942         if ((len > 3) && !strncasecmp(&what[len - 4], ".gif", 4))
 943                 return "image/gif";
 944         else if ((len > 2) && !strncasecmp(&what[len - 3], ".js", 3))
 945                 return  "text/javascript";
 946         else if ((len > 3) && !strncasecmp(&what[len - 4], ".txt", 4))
 947                 return "text/plain";
 948         else if ((len > 3) && !strncasecmp(&what[len - 4], ".css", 4))
 949                 return "text/css";
 950         else if ((len > 3) && !strncasecmp(&what[len - 4], ".htc", 4))
 951                 return "text/x-component";
 952         else if ((len > 3) && !strncasecmp(&what[len - 4], ".jpg", 4))
 953                 return "image/jpeg";
 954         else if ((len > 3) && !strncasecmp(&what[len - 4], ".png", 4))
 955                 return "image/png";
 956         else if ((len > 3) && !strncasecmp(&what[len - 4], ".ico", 4))
 957                 return "image/x-icon";
 958         else if ((len > 3) && !strncasecmp(&what[len - 4], ".vcf", 4))
 959                 return "text/x-vcard";
 960         else if ((len > 4) && !strncasecmp(&what[len - 5], ".html", 5))
 961                 return "text/html";
 962         else if ((len > 3) && !strncasecmp(&what[len - 4], ".htm", 4))
 963                 return "text/html";
 964         else if ((len > 3) && !strncasecmp(&what[len - 4], ".wml", 4))
 965                 return "text/vnd.wap.wml";
 966         else if ((len > 4) && !strncasecmp(&what[len - 5], ".wmls", 5))
 967                 return "text/vnd.wap.wmlscript";
 968         else if ((len > 4) && !strncasecmp(&what[len - 5], ".wmlc", 5))
 969                 return "application/vnd.wap.wmlc";
 970         else if ((len > 5) && !strncasecmp(&what[len - 6], ".wmlsc", 6))
 971                 return "application/vnd.wap.wmlscriptc";
 972         else if ((len > 4) && !strncasecmp(&what[len - 5], ".wbmp", 5))
 973                 return "image/vnd.wap.wbmp";
 974         else
 975                 /* and let xdgmime do the fallback. */
 976                 return xdg_mime_get_mime_type_from_file_name(what);
 977 }
 978
 979 static HashList *IconHash = NULL;
 980
 981 typedef struct IconName IconName;
 982
 983 struct IconName {
 984         char *FlatName;
 985         char *FileName;
 986 };
 987
 988 static void DeleteIcon(void *IconNamePtr)
 989 {
 990         IconName *Icon = (IconName*) IconNamePtr;
 991         free(Icon->FlatName);
 992         free(Icon->FileName);
 993         free(Icon);
 994 }
 995
 996 /*
 997 static const char *PrintFlat(void *IconNamePtr)
 998 {
 999         IconName *Icon = (IconName*) IconNamePtr;
1000         return Icon->FlatName;
1001 }
1002 static const char *PrintFile(void *IconNamePtr)
1003 {
1004         IconName *Icon = (IconName*) IconNamePtr;
1005         return Icon->FileName;
1006 }
1007 */
1008
1009 #define GENSTR "x-generic"
1010 #define IGNORE_PREFIX_1 "gnome-mime"
1011 int LoadIconDir(const char *DirName)
1012 {
1013         DIR *filedir = NULL;
1014         struct dirent *filedir_entry;
1015         int d_namelen;
1016         int d_without_ext;
1017         IconName *Icon;
1018
1019         filedir = opendir (DirName);
1020         IconHash = NewHash(1, NULL);
1021         if (filedir == NULL) {
1022                 return 0;
1023         }
1024
1025         while ((filedir_entry = readdir(filedir)))
1026         {
1027                 char *MinorPtr;
1028                 char *PStart;
1029 #ifdef _DIRENT_HAVE_D_NAMELEN
1030                 d_namelen = filedir_entry->d_namelen;
1031 #else
1032                 d_namelen = strlen(filedir_entry->d_name);
1033 #endif
1034                 d_without_ext = d_namelen;
1035                 while ((d_without_ext > 0) && (filedir_entry->d_name[d_without_ext] != '.'))
1036                         d_without_ext --;
1037                 if ((d_without_ext == 0) || (d_namelen < 3))
1038                         continue;
1039
1040                 if ((sizeof(IGNORE_PREFIX_1) < d_namelen) &&
1041                     (strncmp(IGNORE_PREFIX_1,
1042                              filedir_entry->d_name,
1043                              sizeof(IGNORE_PREFIX_1) - 1) == 0)) {
1044                         PStart = filedir_entry->d_name + sizeof(IGNORE_PREFIX_1);
1045                         d_without_ext -= sizeof(IGNORE_PREFIX_1);
1046                 }
1047                 else {
1048                         PStart = filedir_entry->d_name;
1049                 }
1050                 Icon = malloc(sizeof(IconName));
1051
1052                 Icon->FileName = malloc(d_namelen + 1);
1053                 memcpy(Icon->FileName, filedir_entry->d_name, d_namelen + 1);
1054
1055                 Icon->FlatName = malloc(d_without_ext + 1);
1056                 memcpy(Icon->FlatName, PStart, d_without_ext);
1057                 Icon->FlatName[d_without_ext] = '\0';
1058                 /* Try to find Minor type in image-jpeg */
1059                 MinorPtr = strchr(Icon->FlatName, '-');
1060                 if (MinorPtr != NULL) {
1061                         size_t MinorLen;
1062                         MinorLen = 1 + d_without_ext - (MinorPtr - Icon->FlatName + 1);
1063                         if ((MinorLen == sizeof(GENSTR)) &&
1064                             (strncmp(MinorPtr + 1, GENSTR, sizeof(GENSTR)) == 0)) {
1065                                 /* ok, we found a generic filename. cut the generic. */
1066                                 *MinorPtr = '\0';
1067                                 d_without_ext = d_without_ext - (MinorPtr - Icon->FlatName);
1068                         }
1069                         else { /* Map the major / minor separator to / */
1070                                 *MinorPtr = '/';
1071                         }
1072                 }
1073
1074 //              PrintHash(IconHash, PrintFlat, PrintFile);
1075 //              printf("%s - %s\n", Icon->FlatName, Icon->FileName);
1076                 Put(IconHash, Icon->FlatName, d_without_ext, Icon, DeleteIcon);
1077 //              PrintHash(IconHash, PrintFlat, PrintFile);
1078         }
1079         closedir(filedir);
1080         return 1;
1081 }
1082
1083 const char *GetIconFilename(char *MimeType, size_t len)
1084 {
1085         void *vIcon;
1086         IconName *Icon;
1087
1088         if(IconHash == NULL)
1089                 return NULL;
1090
1091         GetHash(IconHash, MimeType, len, &vIcon), Icon = (IconName*) vIcon;
1092         /* didn't find the exact mimetype? try major only. */
1093         if (Icon == NULL) {
1094                 char * pMinor;
1095                 pMinor = strchr(MimeType, '/');
1096                 if (pMinor != NULL) {
1097                         *pMinor = '\0';
1098                         GetHash(IconHash, MimeType, pMinor - MimeType, &vIcon),
1099                                 Icon = (IconName*) vIcon;
1100                 }
1101         }
1102         if (Icon == NULL) {
1103                 return NULL;
1104         }
1105
1106         /*printf("Getting: [%s] == [%s] -> [%s]\n", MimeType, Icon->FlatName, Icon->FileName);*/
1107         return Icon->FileName;
1108 }
1109
1110 void ShutDownLibCitadelMime(void)
1111 {
1112         DeleteHash(&IconHash);
1113 }